Dynamic user profiles for web personalisation

Highlights

• We describe two algorithms for improving the mapping of interests to an ontology.

• We develop methods for modelling short and long-term user profiles.

• We introduce methods for adapting user profiles based on ongoing user behaviour.

• We demonstrate the effectiveness of our approach in a personalised search system.

Abstract

Web personalisation systems are used to enhance the user experience by providing tailor-made services based on the user’s interests and preferences which are typically stored in user profiles. For such systems to remain effective, the profiles need to be able to adapt and reflect the users’ changing behaviour. In this paper, we introduce a set of methods designed to capture and track user interests and maintain dynamic user profiles within a personalisation system. User interests are represented as ontological concepts which are constructed by mapping web pages visited by a user to a reference ontology and are subsequently used to learn short-term and long-term interests. A multi-agent system facilitates and coordinates the capture, storage, management and adaptation of user interests. We propose a search system that utilises our dynamic user profile to provide a personalised search experience. We present a series of experiments that show how our system can effectively model a dynamic user profile and is capable of learning and adapting to different user browsing behaviours.

Introduction

The dramatic growth of information on the WWW has inadvertently led to information overload and hence finding a specific piece of information has become difficult and time consuming (Challam, Gauch, & Chandramouli, 2007). Web personalisation systems have emerged in recent years in order to deal with this problem aiming to provide a personalised experience to users based on their individual preferences, interests and needs. Such systems have been developed for different domains of application (Pignotti et al., 2004, Challam et al., 2007, Sieg et al., 2007, Pan et al., 2007). In e-commerce, such systems have been used to recommend new items and products to users based on their previous purchasing history (Gorgoglione et al., 2006, Huang et al., 2004), while in e-learning, they are used to provide personalised e-learning services (Sun and Xie, 2009, Zhuhadar and Nasraoui, 2008). Personalisation systems require and maintain information about users, and this may include demographic data, interests, preferences, and previous history. One of the main challenges in such systems is that user interests, preferences and needs are not fixed, but change over time. If user profiles contain just static information, this eventually leads to constraining the personalisation process and recommending irrelevant services and items over time. To overcome this problem, methods are required for learning and understanding different user behaviours, and then adapting the profiles accordingly.

In this paper, we utilise ontological profiles to capture user interests and provide recommendations based on these. The contribution of our work is threefold. Firstly, we introduce two algorithms in order to improve the mapping process between web pages visited by the user that contain implicit information about the user interests and a reference ontology to explicitly represent these interests. Secondly, we introduce novel techniques to construct ontological short-term and long-term profiles that are tailored to the users, and adapt them based on their ongoing behaviour. Thirdly, the methods introduced attempt to recognise and handle potential interest drift and interest shift in the user interests.

To demonstrate our work, we introduce a personalisation system that consists of three phases. The first phase is the information retrieval phase which involves preparing a reference ontology, collecting user navigation behaviour, and mapping visited web pages to the reference ontology. Indeed, this phase is very important as capturing inaccurate user interests would directly affect the subsequent phases and eventually the personalisation performance. In this phase, we utilise two novel algorithms based on our work in Hawalah and Fasli (2011a) to improve the mapping process. The second phase is the profile adaptation and learning phase which utilises previous work in Hawalah and Fasli (2011b). This phase plays a major role in our model as it is responsible for learning, adapting and modelling ontological-user profiles. This also includes methods to adapt the ontological profiles to any shift or drift that might occur in the users’ behaviour. This phase makes use of a multi-agent system that coordinates the various processes and ensures that the user profile remains up-to-date. In the last phase, a re-ranking search system is introduced that utilises the dynamic user profile to provide a personalised search experience. The re-ranking search system takes advantage of the user interests to provide more personalised search results. To evaluate this work, we have conducted experiments with users over a period of time to assess the ability and effectiveness of our methods in tracking and adapting to changes in the user behaviour.

The rest of the paper is structured as follows. First we discuss related work. Section 3 presents the main architecture for modelling dynamic user profiles which consists of three phases with each phase being discussed in more detail in a subsequent section. We introduce the evaluation in Section 4. Section 5 describes the evaluation of the mapping and profile construction methods, while Section 6, details the evaluation of the dynamic user profiling methods in the context of their deployment within a personalised search system. The paper ends with the conclusions and pointers to future work.